A single night’s sleep could hold teh key to predicting your risk for over 100 diseases, according to a groundbreaking new study. imagine knowing your predisposition to conditions like Alzheimer’s or heart failure simply from how you sleep-that’s the promise of this research.
Sleep as a Diagnostic Window
Table of Contents
this study underscores the potential of sleep-based baseline models for risk stratification and longitudinal health monitoring, according to Stanford University co-authors James Zou and Emmanuel Mignot, in collaboration with Rahul Thapa, Magnus Rud Kjaer, Bryan He, Ian Cover, Hyatt Moore IV, Umaer Hanif, Gauri Ganjoo, M. Brandon Westover, Poul and Andreas Bink-kjaer.
Why is Sleep So Important?
With chronic insomnia, obstructive sleep apnea, restless legs syndrome, REM sleep behavior disorder, narcolepsy, delayed sleep phase syndrome, and sleep disorder for shift work among the most common.
A Global Sleep Crisis
Sleep disturbances are widespread. The sleep disorders market is projected to reach $72 billion by 2034,growing at a compound annual rate of 10% between 2025 and 2034,according to Global Market Insights. The American Brain Foundation estimates that 50 to 70 million Americans suffer from sleep-wake disorders. Moreover, about one in three U.S. adults report insufficient rest or sleep daily, according to the U.S. Centers for Disease Control and Prevention. Globally, nearly one billion adults aged 30 to 69 have sleep apnea, as detailed in a 2019 study by Benjafield et al. published in The Lancet Respiratory Medicine.
Decoding Sleep with AI
Researchers developed SleepFM, an AI model trained on polysomnography (PSG) data-data gathered non-invasively during overnight sleep studies. The term “polysomnography,” derived from the Greek prefix “poli” meaning “many,” reflects the recording of numerous physiological signals.
PSG involves painlessly recording brain waves using an electroencephalogram (EEG), blood oxygen levels with pulse oximetry, eye movements via an electrooculogram, heart rate with an electrocardiogram, and breathing and leg movements using an electromyogram. Polysomnography is considered the gold standard for diagnosing sleepwalking, sleep apnea, other sleep-related breathing disorders, chronic insomnia, periodic limb movement disorder, narcolepsy, and REM sleep behavior disorder.
The success of AI models hinges on the quality and quantity of training data. In this study, SleepFM was trained on PSG data from approximately 65,000 participants across multiple cohorts, totaling over 585,000 hours of recordings. These cohorts included data from the Stanford Sleep Clinic (SSC), Outcomes of Sleep Disorders in Older Men (MrOS), the Multi-Ethnic Study of Atherosclerosis (MESA), and BioSerenity, with data from the Sleep Heart Health Study (SHHS) used for algorithm refinement.
“Our model uses 5 to 25 times more data than pretrained supervised sleep or biosignal models,” the researchers wrote.
Using a self-supervised learning algorithm-one that doesn’t require labeled data-the team tested SleepFM against over a thousand disease phenotypes. The AI excelled at predicting Alzheimer’s disease and Parkinson’s disease,both neurodegenerative conditions. Scientists found their AI model accurately predicted 130 conditions, including dementia, stroke, heart failure, chronic kidney disease, myocardial infarction, atrial fibrillation, and all-cause mortality.
Can a single night of sleep truly predict future health risks? this research suggests it can, offering a possibly scalable and efficient method for disease prediction and health monitoring.
“This work demonstrates that base models can learn sleep language from multimodal sleep recordings, enabling scalable and efficient analysis in labeling as well as disease prediction,” the researchers concluded.
